THE PSYCHOMETRIC ASSESSMENT OF OFFENDERS WITH AN INTELLECTUAL DISABILITY MARK KELLS

Transcription

1 THE PSYCHOMETRIC ASSESSMENT OF OFFENDERS WITH AN INTELLECTUAL DISABILITY by MARK KELLS A thesis submitted to The University of Birmingham For the degree of Doctor of Philosophy Department of Psychology Faculty of Science University of Birmingham August 2011

2 University of Birmingham Research Archive e-theses repository This unpublished thesis/dissertation is copyright of the author and/or third parties. The intellectual property rights of the author or third parties in respect of this work are as defined by The Copyright Designs and Patents Act 1988 or as modified by any successor legislation. Any use made of information contained in this thesis/dissertation must be in accordance with that legislation and must be properly acknowledged. Further distribution or reproduction in any format is prohibited without the permission of the copyright holder.

3 ABSTRACT Two psychometric assessments, developed for use with mainstream offenders of normal intelligence, measuring impulsivity and locus of control respectively, were adapted for use with offenders with an intellectual disability (ID). The language and sentence structures used were simplified, and the content of questions was changed to provide contexts that were familiar to detained offenders with ID. Data generated from the responses of 47 male offenders with ID indicated that both tools demonstrated good levels of reliability, in terms of internal consistency, test re-test reliability and correlations with other tools measuring related concepts. Both instruments were found not to differentiate detained offenders with ID (n=47) from two control groups of non-offenders with ID (n=2x46). In addition, an examination of the utility of the impulsivity tool in predicting institutional aggression using receiver operating characteristic (ROC) analysis, was examined. Two measures commonly used for risk assessment, the PCL-R (Hare, 2003) and the HCR-20 (Webster, Douglas, Eaves & Hart, 1997), were also included in this study. The results indicated that the impulsivity measure did not predict institutional aggression. However, the HCR-20 and its sub-scales and the PCL-R and its factors possessed moderate and strong predictive ability respectively, particularly in relation to physical aggression.

4 CONTENTS List of Tables and Figures Acknowledgements Abstract Chapter 1 Background to the thesis Defining intellectual disability Features of intellectual disability Principles of communication Intellectual disability and crime Locus of control and criminal behaviour Impulsivity and criminal behaviour Assessment in intellectual disability function Psychometric assessments of intellectual disability Reliability and validity measurement Retest method Alternative-forms method Split-halves method Internal consistency method Validity of assessments Criterion-related validity Content validity Construct validity The measurement of validity in samples with ID 30

5 1.6. Samples described in the thesis Institutionalised offender sample Institutionalised non-offender sample Non-institutionalised non-offender sample Data collection Structure of thesis Conclusions 36 Chapter 2 A review of response biases implicated in the psychometric assessment of people with an intellectual disability Introduction Acquiescence Negative response bias Suggestibility Critique of suggestibility research Types of assessment Either/or questions Open-ended questions Multiple-choice questions Pictorial assessments Repetition Likert-scales Consistency of Likert-scales Reliability of self-report using Likert-scales Conclusions regarding the use of Likert-scales 56

6 2.5. Possible solutions to alleviate response biases Conclusions 60 Chapter 3 A review examining psychometric tools used in the assessment of offenders with an intellectual disability Introduction Assessment of anger and aggression Assessment of fire-setting Assessment of sexually inappropriate behaviour Pro-offending attitudes Socio-affective functioning Self-management problems Measuring locus of control Measuring empathy Measuring impulsivity Measures of mental health Discussion and conclusion 85 Chapter 4 A revision of the Nowicki-Strickland Internal-External locus of control scale for offenders with an intellectual disability Introduction Locus of control and intellectual disability The measurement of locus of control Aims of research Hypotheses 96

7 4.4. Methods Sample: study 1 and 2a Sample: study 2b Procedure study Procedure study 2a Procedure study 2b Results Study 1: Creation of the M-ANSIE Study 2a: Testing the psychometric properties of the M-ANSIE Study 2b: Factor analysis of the M-ANSIE Discussion Conclusions 119 Chapter 5 A revision of the Eysenck impulsivity questionnaire: an assessment of the validity of self-report impulsivity in a population of offenders with intellectual disability Introduction Impulsivity Impulsivity as a personality characteristic The validity of the psychometric assessment of impulsivity Impulsivity in offenders with an intellectual disability Psychological assessment of people with an intellectual disability Aims of research Hypotheses Methods 135

8 Sample: study 1 and 2a Sample: study 2b Measures Procedure: study Procedure: study 2a Procedure: study 2b Results Study 1: Creation of the I7i-R Study 2a: Testing the psychometric properties of the I7i-R Study 2b: Factor analysis of the I7i-R Discussion Conclusion 161 Chapter 6 The role of impulsivity in the risk assessment of offenders with an intellectual disability in a medium secure hospital setting Introduction Risk assessment in offenders with an intellectual disability Critique of risk assessment in ID The identification of risk factors in samples with ID Dynamic risk factors in offenders with an intellectual disability Psychopathy Personality development in people with an intellectual disability The application of mainstream personality theory Personality disorder in people with an intellectual disability Aims of research 190

9 Hypotheses Methods Sample Measures Procedure Results Multiple regression Receiver operator characteristics analysis Discussion Conclusion 211 Chapter 7 A comparison of self-report impulsivity and locus of control between institutionalised offenders with ID, institutionalised non-offenders with ID and non-institutionalised non-offenders with an intellectual disability Introduction Locus of control Locus of control and criminal behaviour Locus of control and other criminogenic factors Locus of control and institutionalisation Locus of control and offenders in prison Impulsivity and criminal behaviour Aims of research Hypotheses Methods Sample 221

10 Design Measures Procedure Results Discussion Conclusion 226 Chapter 8 Thesis conclusions 228 References 235 List of Tables and Figures Figures Figure 4.1. Error Bar Graph Showing Mean Flesch Reading Ease Statistics for the ANSIE and the M-ANSIE 103 Figure 4.2. Error Bar Graph Showing Mean Item Flesch-Kincaid Grade Level Statistic for the ANSIE and the M-ANSIE 104 Figure 4.3. Error Bar Graph Showing Mean Participant Ratings for Ease of Understanding the ANSIE and M-ANSIE 105 Figure 4.4. Error Bar Graph Illustrating Mean Participant Ratings for Ease of Answering the ANSIE and the M-ANSIE 106 Figure 4.5. Error Bar Graph Illustrating the Mean Number of Items Identified as Difficult to Understand on the ANSIE and the M-ANSIE 107

11 Figure 4.6. Error Bar Graph Illustrating the Mean Number of Items Identified as Difficult to Answer on the ANSIE and the M-ANSIE 108 Figure 4.7. Scatterplot Illustrating the Correlation between the Participants Scores on the ANSIE and the M-ANSIE 110 Figure 4.8. Scree Plot of Eigenvalues for the Principal Components Factor Analysis of the M-ANSIE 113 Figure 5.1. Error Bar Graph Showing Mean Flesch Reading Ease Statistic for the I7i and the I7i-R 142 Figure 5.2. Error Bar Graph Showing Mean Item Flesch-Kincaid Grade Level Statistic for the I7i and the I7i-R 142 Figure 5.3. Error Bar Graph Showing Mean Participant Ratings for Ease of Understanding the I7i and I7i-R 143 Figure 5.4. Error Bar Graph Illustrating Mean Participant Ratings for Ease of Answering the I7i and the I7i-R 144 Figure 5.5. Error Bar Graph Illustrating the Mean Number of Items Identified as Difficult to Understand on the I7i and the I7i-R 145 Figure 5.6. Error Bar Graph Illustrating the Mean Number of Items Identified as Difficult to Answer on the I7i and the I7i-R 146 Figure 5.7. Scatterplot Illustrating the Correlation between the Participant s Scores on the I7i and the I7i-R 149 Figure 5.8. Scree Plot of Eigenvalues for the Principal Components Factor Analysis of the I7i-R 152 Figure 6.1 ROC Curves for Verbal/Property Aggression as Predicted by the I7i, I7i-R and IQ. 202

12 Figure 6.2 ROC Curves for Verbal/Property Aggression as Predicted by the PCL-R, Factor 1, Factor 2 and 13-Item Total. 202 Figure 6.3 ROC Curves for Verbal/Property Violence as Predicted by the HCR-20, H- subscale, C-subscale and R-subscale 203 Figure 6.4 ROC Curves for Physical Violence as Predicted by the I7i, I7i-R and IQ. 203 Figure 6.5 ROC Curves for Physical Violence as Predicted by the PCL-R, Factor 1, Factor 2 and 13-Item Total 204 Figure 6.6 ROC Curves for Physical Violence as Predicted by the HCR-20, H-subscale, C- subscale and R-subscale. 204 Figure 6.7 ROC Curves for High-Risk Violence as Predicted by the I7i, I7i-R and IQ. 205 Figure 6.8 ROC Curves for High-Risk Violence as Predicted by the PCL-R, Factor 1, Factor 2 and 13-Item Total. 205 Figure 6.9 ROC Curves for High-Risk Violence as Predicted by the HCR-20, H-subscale, C- subscale and R-subscale. 206 Tables Table 1.1 Qualitative Descriptions of WAIS-III Full Scale IQ Scores 2 Table 1.2 Guidelines for Interpreting the Size of Correlations 27 Table 1.3 Mean Age, IQ and Length of Time as Resident on the Unit 34 Table 4.1 Ages Appropriate to American School Grades 98 Table 4.2 Examples of the Original and Modified ANSIE Items 99 Table 4.3 Cohen s (1988) Categorisation of Effect Sizes 101 Table 4.4 Readability Statistics for the ANSIE and the M-ANSIE 102 Table 4.5 The Mean Participant Ratings for Ease of Understanding and Ease of Answering the ANSIE and M-ANSIE 105 Table 4.6 Mean Number of Items Identified as Either Difficult to Understand or Difficult

13 to Answer on the ANSIE and M-ANSIE 106 Table 4.8 The Percentage of Externally Oriented Items Endorsed by Yes and No Responses 111 Table 4.9 Eigenvalues and Contribution to the Overall Variance Explained by the 8 Factors 114 Table 4.10 Summary of the Item Content of the Identified Factors 114 Table 5.1 The Primary Traits Comprising the Super-Factors of N, P and E 126 Table 5.2 Summary of Item Content of the Factors of the I7i 129 Table 5.3 Examples of the Original and I7i-R Items 138 Table 5.4 Readability Statistics for the I7i and the I7i-R 141 Table 5.5 The mean participant ratings for ease of understanding and ease of answering the I7i and I7i-R 143 Table 5.6 Mean Number of Items Identified as Either Difficult to Understand or Difficult to Answer on the I7i and I7i-R 145 Table 5.7 The Percentage of Items Endorsed as Impulsive by Yes and No Responses 147 Table 5.8 Means Scores and Standard Deviations Obtained on the I7i and I7i-R 148 Table 5.9 Spearman Correlations of Scores on Self-Report Measures of Impulsivity 150 Table 5.10 Spearman correlations for the self-report measures of impulsivity and the BADS Total and subscales 150 Table 5.11 Eigenvalues and Contribution to the Overall Variance Explained by the 6 Factors 153 Table 5.12 Summary of the Item Content of the Identified Factors 154 Table 6.1 PCL-R 2-Factor, 4-Facet, 20-item Model 176 Table 6.2 Internal Consistency (Cronbach s Alpha) for PCL-R Total, Factor and Facet

14 Scores (Morrissey et al, 2005). 178 Table 6.3 Biserial Correlations Between PCL-R 20- and 13-Item Totals, PCL-R Factors, PCL-R Facets, and Positive and Negative Progress (from Morrissey et al., 2007) 180 Table 6.4 Predictive Accuracy of the PCL-R and HCR-20 in Relation to Positive and Negative Progress Measures: ROC Analysis (from Morrissey, Mooney, et al., 2007) 181 Table 6.5 The Factors and Facets of the Five Factor Model (McCrae, & Costa, 1999) 186 Table 6.6 Means and Standard Deviations for PCL-R 20- and 13-Item Totals, PCL-R Factors, HCR-20 Total and H-, C-, and R-Subscales, and Aggression Types 1, 2 and Table 6.7 Spearman s rho correlations for the independent variables with institutional behaviour exhibited over a 3 month period. 198 Table 6.8 ROC Analyses: ROCs and 95% confidence intervals for three types of aggression occurring within 3 months of assessment. 201 Table 7.1 Mean scores and standard deviations for total scores on the M-ANSIE and I7i-R and in Institutionalised Offenders, Institutionalised Non-Offenders and Non- Institutionalised Non-Offenders. 222

15 CHAPTER ONE: Background to the Thesis The aim of this thesis was to advance our understanding of offending and the assessment of offending behaviour in people with intellectual disability (ID). In addition, an attempt was made to generate findings that would be of practical use by those employed in the criminal justice system (CJS). This has been done by demonstrating the adaptation of two valid and reliable tools measuring self-report locus of control (LOC) and impulsivity respectively. This thesis has attempted to add to our empirical knowledge of response biases and the adaptation of assessment tools to minimise the threat of these response biases, and thus to improve the reliability and validity of self-report measures used with offenders with ID. Empirical knowledge has also been gained in relation to the predictive accuracy of two tools used in risk assessment, the Psychopathy Checklist - Revised (PCL-R: Hare, 2003) and the Historical Clinical Risk-20 (HCR-20: Webster, Douglas, Eaves & Hart, 1997), developed in relation to mainstream non-id offenders, when used with offenders with ID. The thesis has also sought to clarify whether two factors associated with offending in mainstream non-id samples, external orientation of LOC and impulsivity, are also elevated in a sample of offenders with ID. This chapter consists of a definition of ID and an outline of some the deficits associated with ID that may impact upon self-report assessment in people with ID. The literature relating to the prevalence of people with ID in the JCS is then critically reviewed and summarised. The evidence relating LOC and impulsivity to offending behaviour is also critically reviewed and summarised. The need to assess the reliability and validity of measures is discussed and material relating to different forms of reliability and validity is presented. The actual composition of samples in studies concerning people with ID is critically reviewed. The sample for this thesis and the data collection process are then described. The structure of the thesis is then outlined. 1

16 1.1. Defining Intellectual Disability Intellectual Disability (ID) is defined as a significant impairment of intelligence and social functioning acquired before adulthood. The three core criteria within this definition have gained widespread acceptance across professional boundaries in both the UK and America (American Association on Mental Retardation, 1992; American Psychiatric Association, 2000; British Psychological Society, 2001; Royal College of Psychiatry, 2001; World Health Organisation, 1992), these criteria are: Significant impairment of intellectual functioning (IQ <70); Significant impairment of adaptive/social functioning (significantly reduced functioning in at least two of the following areas: communication; self-care; home living, social/interpersonal skills; use of community resources; self-direction; functional academic skills, work, health and safety); Age of onset before adulthood (< 18years old). Significant impairment of intellectual functioning is determined through the administration of a reliable, valid and properly standardised psychometric assessment, such as the Wechsler Adult Intelligence Scales, currently its fourth edition (WAIS-IV: Wechsler, 2010). The WAIS-III was used for all assessments of intellectual functioning in this thesis. The mean IQ score on the WAIS instruments is 100 with a standard deviation of 15. About two thirds of all scores on the WAIS fall within one standard deviation of the mean (IQ range= ), and about 95% fall within two standard deviations of the mean (IQ range=70-130) (Wechsler, 1998). Significant impairment in intellectual functioning is defined as an IQ below the lower boundary of two standard deviations from the mean, represented by an IQ of less than 70. Impairment of adaptive/social functioning is determined by the use of one of several scales designed for that purpose, the best known of which is the Vineland Adaptive Behaviour Scale 2

17 (VABS: Sparrow, Balla & Cicchetti, 1984) and are recommended by the British Psychological Society (BPS) (2001) and the World Health Organisation (WHO) (1992). Table 1.1 shows the qualitative descriptions of the WAIS-III full IQ scores. Table 1.1. Qualitative Descriptions of WAIS-III Full Scale IQ Scores IQ Score Classification Theoretical Percentage of population within IQ range 130 and above Very Superior Superior High Average Average Low Average Borderline and below Extremely Low Features of ID The performance of an individual with ID during an assessment will be mediated by motivational and cognitive processes (Beirne-Smith, Patton & Kim, 2006). Motivational factors include self-determination, mediated by locus of control (which will be discussed in detail in Chapter 4), outer-directedness and expectancy of failure (Luthar & Zigler, 1988; Yando & Zigler, 1971; Zigler, 1961; Zigler, Bennett-Gates, Hodapp & Henrich, 2002; Zigler & Burack, 1989), as discussed in Chapter 6. Cognitive processes include attention (Krupski, 1980; Krupski, 1985), memory (Brown, 1972; Brown, Campion & Murphy, 1974; Brown & Scott, 1971) and communication difficulties (Abbeduto & Nuccio, 1991). Locus of control refers to the extent that people view events in their life as controlled by themselves or others. People with an external locus of control tend to view life s events as controlled by others (Nowicki, 1976). Outer-directedness refers to a problem-solving strategy that relies upon 3

18 external cues for guidance (Balla & Zigler, 1979). Evidence suggests that people with ID have an external locus of control (LOC) (Wehmeyer & Palmer, 1997) (see Chapter 4) and tend to adopt outer-directedness as a problem-solving strategy (Balla & Zigler, 1979) (See Chapter 6). Research has also demonstrated that people with ID have a high expectancy of failure (Cummins & Das, 1980; Schloss, Alper & Jayne, 1994). The detrimental effects of repeated failures are consistent with all three of these motivational themes (Zigler et al., 2002). It would seem reasonable to assume therefore, that the net result of these motivational difficulties is that many people with ID may be prone to under-perform in a variety of social and academic contexts. Evidence presented in Chapter 2 suggests that people with ID do not always perform well using a variety of questioning techniques when compared to peers without ID (Gudjonsson & Henry, 2003) Deficits in attention (Krupski, 1980; Krupski, 1985), short-term memory (Brown, 1972; Brown et al., 1974; Brown & Scott, 1971) and long-term memory (Cantor & Ryan, 1962), and receptive (Abbeduto et al, 1989; Abbeduto & Nuccio, 1991; Bartel, Bryen & Keehn, 1973; Cromer, 1972, 1974) and expressive language (Dunn & Dunn, 1981), have been reported in populations with ID when compared with people of normal IQ. However, these are not global deficits. Participants with ID perform as well as participants with normal IQ on simple tasks testing attention (Blackwell, McIntyre & Murray, 1983; Krupski, 1985), memory (Brown, 1972) and language skills (Abbeduto et al., 1989). However, as tasks become more complex performance deteriorates. For example, the Span or Apprehension Task is designed to measure selective attention (Krupski, 1987). In this task, participants have to spot target letters displayed on a screen when embedded within a series of interference letters. When embedded within 0, 2 or 4 interference letters no differences between participants with or without ID were observed. However, when the interference 4

19 series extends to eight letters or more, participants with ID perform significantly worse than matched participants of normal IQ (Blackwell et al., 1983). However, there are problems associated with these studies, which include: small sample sizes (Abbeduto et al., 1989; Abbeduto & Nuccio, 1991; Brown, 1972); lack of specificity regarding the reporting of IQ for ID groups (Blackwell et al., 1983; Krupski, 1980; Krupski, 1985); and (possibly due to the developmental nature of this work) an overrepresentation of studies including children as participants (Abbeduto et al., 1989; Abbeduto & Nuccio, 1991; Blackwell et al., 1983; Krupski, 1980; Krupski, 1985). These limitations mean that caution should be exercised when generalising from the results. Regardless, the conclusions of studies overwhelmingly indicate deficits in these areas associated with ID. These motivational and cognitive difficulties are likely to impact upon the ability of people with ID to answer questions. For example, attention deficits are likely to make any interview process more taxing and place higher demands upon the cognitive mechanisms of a person with ID. Long-term memory difficulties may hinder the recall of events. Short-term memory difficulties may prevent the person from holding the question in their memory while they decide upon an appropriate response. Difficulties with receptive language mean that an individual with ID may be unable to understand questions, particularly as those questions become more complex. Difficulties with expressive language may prevent an individual with ID from being able to express their actual attitudes or beliefs Principles of Communication In view of these difficulties, Mencap (2010) has produced guidelines on how to provide written information to people with intellectual disabilities. However, the same principles apply to spoken communication. The basic principles are: give some thought to the questions you will need to ask before the interview starts; 5

20 keep the language clear and simple; try to keep to one idea per sentence; keep sentences short; try to use the same word for the same thing; find out what the person s own words are for specific things (this can often be culturally based); be supportive. In addition, Bull (2010) summarises the factors that interviewers need to consider when speaking to vulnerable witnesses, such as people with ID, in order to gain a valid account of an incident. It is likely that these factors will help to facilitate valid responding in wider contexts. Interviewers need to: speak more slowly; allow extra time to enable the person with intellectual disabilities to take in what is being said; allow time for the person with intellectual disabilities to think about how they are going to answer the question; do not rush the questions; avoid interrupting; be patient. The interviewing procedures used in gathering data for this thesis conformed to these guidelines. When interviewing an individual as part of an assessment it is important that the information gained is reliable and valid. 6

21 1.3. ID and Crime Ascertaining the true nature and prevalence of offending in ID populations is a complex and problematic venture not least because research does not occur within a socio-cultural vacuum. Pressures associated with public concern about the threat presented by certain groups, political policy, the wider socio-cultural context, the imprecise nature of definitions and measurement of ID, attrition and filtering at all levels of the criminal justice system and the consequent ambivalence of crime statistic are active within the processes and the implications of research (Day, 1994; Holland, 2004; Holland, Clare & Mukhopadhyay, 2002; Lyall, Holland, & Collins, 1995a; Murphy & Mason, 1999). In addition, those research studies considering the prevalence of offending in ID populations are methodologically flawed. It is not within the scope of this paper to consider all of those methodological difficulties (for a discussion of these issues see Holland, 2004) but rather to provide a flavour of the findings. As attrition occurs at many points between the commissioning of a potentially illegal act and conviction, the end-point of the criminal justice process, it is important at what stage of the CJS process sampling takes place (Holland, 2004). Consequently, it is likely that prevalence will be lower in the later stages of the CJS due to diversion policies (Simpson & Hogg, 2001). Some studies have highlighted the differential responses of professionals in the criminal justice system when dealing with people with ID (Cant & Standen, 2007; McAfee, Cockeram & Wolfe, 2001) and some have also included the interface between the CJS and community services (Fryson, 2005; Keilty & Connelly, 2001; Lyall, Holland, Collins & Styles, 1995; McBrien & Murphy, 2006). Several of these studies highlight the reluctance of care workers or teachers to report crimes to the police, citing reasons such as fear of blame or investigation of their own practices and the labelling of the perpetrator (Fryson, 2005; Keilty & Connelly, 2001; Lyall, Holland, Collins & Styles, 1995). For example, McBrien and 7

22 Murphy (2006) reported that only 48% of carers included in their study would report theft, 68% would report an assault and 83% would report a rape. Fryson (2005) reported that 88% of schools had experienced pupils behaving in a sexually inappropriate way, in 65% of schools this occurred at least once a term and in 19% of schools it occurred on a weekly basis. Of the incidents reported 58% involved public masturbation and 15% involved actual or attempted bodily penetration. In response to incidents, 54% of schools had approached social services but only 23% had approached the police. In addition, it has been reported that the police are reluctant to investigate crimes committed by people with ID, preferring incidents to be dealt with by care staff (Keilty & Connelly, 2001). However, McBrien and Murphy (2006) in a more recent study reassuringly report that all of the police officers included in their study felt that rape and assault should be reported, while 72% felt that a theft should be reported. McAfee et al. (2001) report police officers differential responses to crime scenarios based upon the presence or absence of ID in the perpetrator or victim of a crime. Officers reported that they were more likely to believe the victim of crime if they had ID and were more likely to provide extra help for a victim with ID. For minor crimes, officers reported stronger action against a perpetrator when the victim had ID but weaker action against a perpetrator with ID. For major crimes officers were less likely to believe the victim when the perpetrator had ID but were more likely to rate the crime as serious. These studies highlight some of the complexity at the interface of the community and the CJS. The failure to report offences or to investigate potential offences will clearly impact upon the number of people with ID progressing into the CJS. It should be noted however, that there are limitations to the studies described above. For example, all include relatively small sample sizes, with the exception of McBrien & Murphy (2006), which included 80 care staff and 65 police officers. In the study citing the figures of most concern (Lyall, Holland, 8

23 Collins & Styles, 1995) it is unclear how the managers of care-homes were interviewed, or indeed how many were interviewed, and consequently it is difficult to generalise from the findings. In addition, the reliability of offending rates is not clear in all studies, either because offences are identified by carers rather than actual crime reports (Lyall et al, 1995) or because the definition used, for example, sexually inappropriate behaviour may include behaviours over and above those that are offences (Fryson, 2005). Other studies are based upon paper and pencil exercises in which a number of variables are manipulated (McAfee et al., 2001; McBrien & Murphy, 2006). It is difficult to gauge the extent to which responses to hypothetical scenarios generalise to actual behaviour in reality. Despite filtering of potentially criminal acts described above, many offences involving suspected offenders with ID are investigated by the police. Data suggest that between 0.9% and 9% of people interviewed in police stations have an IQ below 70 (Irving, 1980; Irving & McKenzie, 1989; Lyall, Holland & Collins, 1995). Lyall et al. (1995) identified 5% of people taken into police custody in a single police station over a three-month period as having ID. Winter, Holland and Collins (1997) screened individuals leaving or entering police custody over a 33 day period. Formal IQ testing identified two people (1%) with IQs below 70. However, there are problems with these studies. Both studies relied upon the use of a screening tool (Clare & Gudjonsson, 1992) to identify potential participants. Clare and Gudjonsson (1992) identified problems with the screening tool. In their study it led to the identification of 5% false positives, where two participants from 39 identified themselves as having reading difficulties or a learning disability which subsequent formal testing indicated was not ID. Equally, it led to the identification of 20% false negatives, where 11 from 54 participants who did have ID, did not identify themselves. The net effect is that the screening tool is likely to result in a net under identification of people with ID. Consequently, the failure of Lyall, Holland and Collins (1995) to complete a formal 9

24 IQ assessment is likely to result in the under identification of people with ID. In the Winter et al. (1997) study, only 21 of the 47 people identified as possibly ID agreed to participate in the study and therefore it is possible that some people with ID were not tested, leading to a lower identification rate. In the Lyall, Holland and Collins (1995) study, the screening tool was only used with 25% of those entering custody. No explanation was given for this but this non-random sampling may have affected the results. Some studies have considered the vulnerability of suspects with ID in police custody (Clare & Gudjonsson, 1995; Leggett, Goodman & Dinani, 2007). These studies highlight the lack of understanding of the consequences of police interviewing and the use of interviewing techniques not suited to gaining reliable and valid information. No British studies could be identified that have attempted to assess the prevalence of ID in relation to court appearances. However, a small scale qualitative study again highlights the different attitudes to dealing with people with ID in these contexts (Cant & Standen, 2007), some of which may again lead to filtering away from formal legal procedures. However, two Australian studies identified 24% and 10% respectively, of people accused of committing offences appearing in four Australian courts as having IQs below 70, and a further 14% and 20% respectively, had IQs in the borderline range (Hayes, 1997; Vanny, Levy, Greenberg & Hayes, 2009). There are several methodological problems with these studies. Both studies adopt a non-random selection method and use samples not representative of the wider Australian population. In addition, IQ was determined using the Matrices Section of the Kaufman Brief Intelligence Test (K-BIT: Kaufman & Kaufman, 1990) in the first study, and the K-BIT in the second, although it is not clear if the full assessment was used. The reliability and validity of this measure and the procedure used to estimate the full-scale IQ is not clear. In addition, it is unclear whether cultural differences of 10

25 indigenous Australians, included in the study, may have impacted on their scores, thus underestimating actual intellectual ability. If conviction occurs offenders with ID may face incarceration. The mean IQ score for convicted prisoners is below average suggesting that many prisoners are intellectually disadvantaged (Birmingham, Mason & Grubin, 1996; Eysenck & Gudjonsson, 1989; Mason & Murphy, 2002). Consequently, one might expect high numbers of offenders with ID in the prison population. However, studies generally report low levels of people with IQs less than 70 in prison populations (Fazel, Xenitidis & Powell, 2008). For example, Murphy, Harnett and Holland (1995) reported 0%. However, Birmingham, et al. (1996) reported 13% and Crocker, Cote, Toupin and St-Onge (2007), using a Canadian sample, report that 19% of their sample were in the probable ID range, with a further 30% in the borderline range of intelligence. Sondennaa, Rasmussen, Palmstierna and Nottestad (2008) using a Norwegian prison sample, report a prevalence rate of 11% with IQs below 70 and an additional 20% in the borderline range. Holland and Persson (2010) assessed prevalence at 1% using an Australian prison sample. These figures suggest a wide range of prevalence rates in prisons and highlight the extremely large numbers of prisoners in the borderline range. However, methodological problems make comparisons difficult and make generalising from the results problematic. Two of the studies considered men on remand and therefore, were not offenders and were unlikely to be representative of the wider prison population (Birmingham et al., 1996; Murphy et al., 1995). The means of determining IQ was different in every study and the reliability of those measures is questionable in several, using screening tools, not specifying the measure used, or reliance upon official records, without reference to how IQ was actually assessed (Birmingham et al., 1996; Crocker et al., 2007; Holland and Persson, 2010; Murphy et al., 1995). In addition, for a number of reasons, only 38% (Crocker et al., 2007) and 77% (Sondennaa et al., 2008) of the 11

26 originally identified samples were used. Consequently, it is possible that some non-random sampling had affected the results in these studies. Shelton (2006) reported the prevalence of ID in a juvenile justice system for offenders between the ages of 15- and 17-years in the US at 20%. This study included a good sample size, containing 290 (25%) of all male detainees and 60 (100%) of female detainees over a four-month period. However, it is unclear how the level of ID was determined, as the means of determining IQ was not specified and no reference was made to the level of adaptive functioning of the sample. In addition, individuals with IQs as high as 82 were included in the ID group, suggesting that the level of ID was over-estimated, as these IQs actually fall in the low/average range. Mason and Murphy (2002) report the prevalence of ID in a sample under the supervision of the probation service. Mason & Murphy (2002) report WAIS-R Short-Form (WAIS-R; Wechsler, 1981) and VABS (Sparrow et al., 1984) scores, and define ID as individuals who scored two standard deviations below the mean on both assessments. The authors reported a prevalence rate of 6% with IQs of 75 or less (allowing for two standard errors). However, if just the WAIS-R scores were used to determine intellectual functioning the figure rose to 17%. The findings of this study are limited by a small sample size (n=70) and a non-random sampling technique, based on the case-load of probation officers who volunteered to take part. Another research strategy for identifying the prevalence of offending in ID populations is to consider the wider population of people with ID and then attempt to assess how many of them have had contact with the CJS. Studies have identified between 1% and 10% respectively of people known to UK developmental disability services having been in contact with the police or other criminal justice agencies because of alleged offending (Lyall, 12

27 Holland & Collins, 1995; McBrien, Hodgetts & Gregory, 2003; McNulty, Kissi-Deborah & Newsome-Davies, 1995; Seaward & Rees, 2001). Studies have reported that 2%, 10%, 5% and 1% respectively, of their samples had been in contact with the CJS in relation to the perpetration of a crime (Lyall, Holland, Collins & Styles, 1995; McBrien et al., 2003; McNulty et al, 1995; Seaward & Rees, 2001). In addition, Tyrer et al. (2006) reported on a large sample (n=3062) drawn from the Leicestershire Learning Disability Register. They reported that 14% people with had been physically aggressive towards others in the preceding two years. In this instance physical aggression was determined from carers responses to a questionnaire routinely completed as part the services assessment process. In relation to 80 (3%) individuals, the level of physical aggression had resulted in serious injury to the other person. In response to this aggression, 17 individuals (1%) had been permanently excluded from day services and only 45 (1.5%) resulted in the police taking or threatening to take formal action. Cooper et al. (2009) used a two year longitudinal study to identify the prevalence of aggressive behaviour within a geographically defined area of Scotland. They report prevalence of aggression (including verbal aggression, destructive aggression and physical aggression) as 10%. In addition, the remission rate, that is to say those who had been violent prior to the start of the study and were violent again during the two year follow-up, was 28%. The advantage of these studies is that they often included large sample sizes (Lyall, Holland & Collins, 1995 (n=358); McBrien et al., 2003 (n=1326); Tyrer et al., 2006 (n=3062)). However, generalisation from these studies is problematic as they do not utilise representative samples, only including those accessing or known to services. Inevitably the sample only includes those who agree to participate, which may influence prevalence rates if those displaying aggressive behaviours are more reluctant to participate than those not displaying aggressive behaviour. None of the studies assessed ID independently, instead 13

28 relying upon the criteria for access to services in the particular locality. Consequently, it is unclear what range of ID is included, although the Lyall, Holland and Collins (1995) study appears to include those in the borderline range. A major limitation of these studies, however, is the identification of contact with the CJS and the identification of offending type behaviour. This information is often gained from service managers or care workers and therefore its validity in terms of actual offending rates is uncertain (Lyall, Holland & Collins, 1995; McBrien et al., 2003; McNulty, et al., 1995; Seaward & Rees, 2001; Tyrer, et al., 2006). Cooper et al. (2009) gathered information from a wide variety of sources in order to tackle this issue, including case-notes, and interviews with participant, carers and family members where possible. However, no evidence of the validity of this procedure was reported. There is also evidence of non-random sampling and other problems with the data sampling procedures. For example, Tyrer et al., (2006) report that only 50% of individuals on the Leicestershire Learning Disability Register have had an interview and that for those that have had interviews there are updates every five to seven years. Consequently, the range of participants that could be included in the study is restricted and the data accessed could be out-dated. Using a different methodology, Dickson, Emerson and Hatton (2005) conducted a secondary analysis of data relating to self-reported antisocial behaviour, obtained from the 1999 Office for National Statistics (ONS) study of the mental health of children and adolescents in the UK. The authors reported that 16% of the group with ID self-reported being in trouble with the police compared to 8% of the group without ID. This study utilised a large sample size (n=4174) of whom (n=124), 3% were identified as having ID. However, it is unclear how ID was determined in this analysis and the validity of the self-report is uncertain in relation to the actual commissioning of a crime. Equally, there is evidence of 14

29 non-random sampling as information was only available for 79% of the sample with ID and 94% of the sample without ID. The literature regarding prevalence rates is equivocal (Simpson & Hogg, 2001). Methodological difficulties and procedural differences make it difficult to draw firm conclusions from the available evidence. In an attempt to overcome those methodological problems Hayes, Shackell, Mottram and Lancaster (2007) report on a well controlled study which assessed the prevalence of ID in a large big city local prison. A random sample of 140 prisoners, representing approximately 10% of the prison population, participated in the study. The WAIS III-UK Version (WAIS III (UK), Wechsler, 1999) was used to determine IQ. Vineland Adaptive Behaviour Scales Interview Edition (VABS, Sparrow et al., 1984), completed using self-report, was used to determine the level of adaptive behaviour. On the WAIS-III (UK), 7% of participants achieved standard scores below 70 and a further 24% were in the borderline intelligence range. Therefore, a total of 31% showed considerable deficits in intellectual abilities. Some community services include clients with standardized scores of up to 75 on psychometric grounds (Holland, 2004). Consequently, this group was also analysed. Results indicated that 16% of the sample had a score of 75 or less on the WAIS-III. The VABS results indicated that 10% of participants had standard scores below 70 and a further 33% were in the borderline range yielding a total of 43% with serious deficits in adaptive behaviour. In addition, 26% had a score of less than 75. These findings suggest that the number of people with ID in the UK prison is larger than previously determined. However, only four participants (3%) had standard scores below 70 on both the VABS composite standard score and WAIS-III (UK) Full Scale IQ (FSIQ), with thirteen participants (9%) falling below 75 on both instruments. When the borderline groups (<80) on both tests were included, a total of 30 participants (22%) fell below 79 on both assessment instruments. Whilst there are differences in the pattern of distribution between the VABS 15

30 and WAIS-III, the two tests correlate significantly (r=.57). The mean total scores on the WAIS-III (87.09, SD=12.5) and the VABS (82.51, SD=13.2) were both in the low average range. This study represents a clear attempt to assess the level of ID in a UK prison with regard to IQ and adaptive functioning. However, even in this reasonably well controlled study the continuing debate over the definition of ID is apparent. Using a strict 70 demarcation, 3% of offenders had ID. However, allowing for the 2 standard errors often utilised in ID services to demarcate admission criteria, 9% scored below 75 on both measures, whereas including the borderline range increased the percentage to 22%. These figures suggest that the number of people with intellectual deficits in UK prisons is higher than previously supposed. However, caution is required in generalising from this study. With only 140 participants, this is a relatively small sample size with which to generalise to the wider prison population. It is also unclear to what extent this sample was representative of the wider prison population; the authors note that black and minority ethnic groups were under-represented. In addition, the sample included those on remand and therefore, strictly speaking, is not entirely composed of offenders as some may not be found guilty. In addition, 39 offenders refused to participate, 18 withdrew during data collection and seven were transferred or released before data collection was completed. It is unclear whether this substantial group, 31.37% of the original 204 identified, had any characteristics in common which might have influenced the results. In addition, the authors did not report on the reliability of the VABS assessments. Without, a reliable assessment of social/adaptive functioning it is possible that researchers will continue to define ID primarily in terms of IQ. However, Talbot and Riley (2007) in a qualitative study noted that some individuals, although aware of their intellectual limitations, had learnt to hide them under a cloak of 16

31 competence for fear of discrimination. Consequently, some individuals with ID may refuse to participate in research such as this in order to avoid identification. It is difficult to draw firm conclusions about the prevalence of people with ID in the CJS. Some differences may be due to sampling at difference stages of the CJS, as some authors have suggested (Simpson & Hogg, 2001). However, such a pattern was not obviously apparent in the evidence presented above. This may have been due to the significant methodological differences between studies and particularly differences in the means of determining the level of ID of the sample. Some have sampled during the early stages, such as people detained in police stations (Irving, 1980; Irving & McKenzie, 1989; Lyall, Holland and Collins, 1995) and others at late stages, such as prison populations (Crocker et al., 2007; Hayes et al., 2007; Murphy et al., 1995). Some seek to identify the numbers of people with ID present in offender populations (Hayes et al., 2007) and other seek to identify how many individuals known to ID services have been involved with the CJS (McBrien et al., 2003; McNulty, et al., 1995; McBrien, et al., 2003; Seaward & Rees, 2001; Tyrer, et al., 2006). Many use small sample sizes (Fryson, 2005; Keilty & Connelly, 2001; McBrien & Murphy, 2006) and others large sample sizes (Cooper et al., 2009; Tyrer et al., 2006). Some use abbreviated tests to assess level of ID (Mason & Murphy, 2002), others use screening tools (Lyall, Holland & Collins, 1995; Winter et al., 1997) and others use more robust measures such as the WAIS (Hayes et al., 2007). This is often inversely related to sample size with larger sample sizes often relying upon quicker, less valid measures or the identification of ID by the service being examined (Cooper et al., 2009; Tyrer et al., 2006) and smaller sample sizes utilising more robust measures (Mason & Murphy, 2002). Hayes et al. (2007) may be considered an exception with full WAIS and VABS assessments for a sample of 120 offenders. However, even this sample may be considered small when used as an estimate of the level of ID in the whole prison population, of around offenders 17

32 (Ministry of Justice, 2010). In addition, few studies incorporate assessment of adaptive functioning in a formal sense, although Hayes et al. (2007) was an exception. However, even in that instance the validity of the VABS was uncertain due to the use of a self-report version, rather than administration to a third-party as designed (Sparrow et al., 1984). Many researchers and clinicians have expressed opinions regarding the potential overor under-representation of ID offenders commissioning specific offences (Barron, Hassiotis & Banes, 2004; Hayes 1993; Hayes & Craddock, 1992; Holland, 2004; Lindsay, 2002; Lindsay, O Brien, Carson, Holland, Taylor, Wheeler.,...Johnston, 2010; Simpson & Hogg, 2001; Walker & McCabe, 1973). Studies conducted in different contexts tend to produce different prevalence rates and different profiles of offences committed (Hogue et al., 2006; Lindsay, Hastings, Griffith & Hayes, 2007). For example, in terms of more recent studies, Dickson et al. (2005) reported that adolescents with ID self-reported higher levels of bullying/threatening others; stealing valuable items from houses/ shops/school; using weapons against others; starting fires; deliberately destroying property; stealing from someone in the street, when compared to those without ID. In contrast, Lindsay, O Brien, et al. (2010) assessed the offence histories of those in offender services for people with ID and reported that fire setting, theft and road traffic offences do not feature prominently. The most frequent index offences and prior problems related to physical and verbal aggression. Holland and Persson (2010) report that prisoners with an ID were younger than offenders without ID, at the commencement of their current incarceration and at their first term of incarceration. In terms of the most serious offence (MSO) for which prisoners were serving their current term of imprisonment, prisoners with an ID were significantly more likely to have a property offence and significantly less likely to have a drug offence as their MSO than prisoners from the non-id sample. In terms of those purely in high secure settings, Hogue et al. (2006) reported significantly more convictions for threatening behaviour, assault 18